Probabilistic modeling of ship powering performance using full-scale operational data

The energy efficiency of ocean-going vessels can be increased through various operational considerations, such as improved cargo arrangements and weather routing. The first step toward the goal of maximizing the energy efficiency is to analyze how the ship's powering performance changes under different operational settings and weather conditions. However, existing analytical models and empirical methods have limitations in reliably estimating the powering performance of full-scale ships in real operating conditions. In this study, machine learning techniques are employed to estimate the powering performance of a full-scale ship by constructing regression models using the ship's operational data. In order to minimize the risk of overfitting in the regression process, domain knowledge based on physical principles is combined into the regression models. Also, the uncertainty of the estimated performance is evaluated with consideration of the environmental uncertainties. The obtained regression models can be used to predict the ship speed and engine power under different operational settings and weather conditions.     

一种基于蓝牙室内指纹定位的贝叶斯改进算法

贝叶斯估计是重要的位置指纹定位算法,但传统的等值贝叶斯先验概率在动态定位中不适用。针对该问题,本文提出了一种基于贝叶斯指纹定位的改进算法。首先,借助陀螺仪获取的航向信息和高斯核函数模型建立概率投票算法,计算先验概率;然后,结合先验概率和信号强度计算待测点位于参考点上的后验概率;最后,选取概率最高的参考点,以概率为权重计算待测点的最或然值。以智能手机为试验对象,在规则路径试验中,改进算法的平均定位误差为1.15 m,定位误差小于2 m的概率为96.1%,不规则路径试验中,平均定位误差为0.50 m,定位误差在1 m的可信度为94.8%;并且改进算法对定位中位置跳变的现象有明显改善,具有较好的稳健性。     

弹性各向同性介质一次散射波场高斯束Born正演

本文提出了一种弹性波一次散射波场的正演方法——弹性波高斯束Born正演.该方法以线性散射理论为基础,通过Born近似建立起地下散射点处不同波型的反射率同弹性波主分量波场之间的数据映射关系,利用高斯束所包含的走时、振幅和极性信息进行不同波型的局部平面波的合成,进而通过逆倾斜叠加将所合成的局部平面波转化为时空域的多分量地震记录.该方法不但保持了射线类方法高效的优点,还具备了处理多次走时波场的能力,从而保证了复杂构造的波场模拟的精度.文中两个数值模型的应用效果表明,本文所提出的弹性波高斯束Born正演算法具有近似于波动方程有限差分法的波场模拟精度以及高得多的计算效率.     

基于黏声介质的角度域高斯束逆时偏移方法研究

高斯束逆时偏移是一种兼具计算效率和成像精度的深度域成像方法,能够面向目标成像.地下介质中黏滞性普遍存在,利用传统各向同性或完全弹性的成像方法处理黏滞性探区的数据会降低分辨率,并导致成像位置不准确和振幅欠估计等问题.本文在高斯束逆时偏移的基础上,通过对震源点和检波点处的波场进行衰减补偿,并结合高斯束求解时的角度信息,实现了黏声介质角度域高斯束逆时偏移方法.最后通过模型和实际资料试算对本文方法的正确性和适用性进行了验证.试算结果表明:相比于声波高斯束逆时偏移,本文方法能够对黏滞性引起的吸收衰减进行有效补偿,同时提取的角度域共成像点道集(ADCIGs)不仅可以用于分角度叠加成像压制成像噪声,而且能够为后续的偏移速度分析提供支撑.     

结合序贯指示模拟的贝叶斯高斯混合线性反演

高斯混合模型(Gaussian Mixture Model, GMM)可以用来描述储层性质的多峰分布特性,多峰特性主要是由于它们在不同离散变量内的变化而引起的.在高斯混合模型中,高斯分量的权值代表离散变量的概率.然而,基于高斯混合模型的贝叶斯线性反演可能会对某些点的离散变量错误地分类,进而影响连续变量的反演结果,尤其存在强噪声的时候.在本文中,我们考虑了离散变量的空间变化性,并将高斯混合模型与序贯指示模拟(Sequential Indicator Simulation, SIS)相结合来确定离散变量的后验条件权值,形成了结合序贯指示模拟的贝叶斯高斯混合线性反演方法.该方法能够准确地对离散变量进行归类,且具有良好的抗噪性.通过模型试算,我们证明了这种方法的可行性,并在实际资料中取得了较好的结果.     

共炮检距高斯波束偏移的一种快速实现算法

与共炮高斯波束偏移相比,共炮检距高斯波束偏移具有直接抽取炮检距域共成像点道集的优势.过去,共炮检距高斯波束偏移以损失成像精度的代价采用最速下降法来降低积分的维数,从而提高计算效率.但经过最速下降近似简化的偏移公式仍是频率域的,需要在每个频点进行计算.为此,本文提出一种快速实现算法来避免采用最速下降法.本文通过分析一个水平层状速度模型的偏移过程和Marmousi速度模型的成像结果来检验不同插值方法对快速实现算法的成像精度和计算效率的影响,并建议采用二维三次卷积插值方法.同时本文在Marmousi速度模型下验证了快速实现算法相对于最速下降法在成像精度和计算效率上的优势.此外,本文将采用二维三次卷积插值的快速实现算法应用于Sigsbee2A模型并获得了清晰的盐下图像.     

高斯波包反射走时速度反演方法

扰动高斯波包理论指出,在Gabor域描述模型的扰动成分,且入射波场为短时宽带信号时,扰动波场可在时间域通过高斯波包算子描述.在此基础上通过拟合反射波的走时,提出一种速度反演方法.反射波走时残差利用地震道局部波形的互相关函数表示,以走时残差的二范数作为目标函数,优化目标函数实现对速度场的反演.基于一阶Born近似,利用扰动高斯波包理论推导出目标函数对速度场的梯度是本文理论部分的核心内容.梯度包括两部分:正传的背景波场与反传的扰动高斯波包之间的互相关,反传的背景波场和正传的扰动高斯波包之间的互相关.梯度表达式中背景波场和扰动波场均利用高斯波包算子模拟.计算梯度的具体算法中,如何模拟扰动波场,以及如何计算反射波的走时残差是两个要点,文中对此做了详细的讨论.数值实验进一步阐述了反演的实现策略,实验结果表明高斯波包反射走时速度反演方法和实现策略有效可行,并得到了理想的反演结果.     

Field observations of the vertical distribution of sand transport characteristics over fine,medium and coarse sand surfaces

The vertical distribution of sand transport characteristics is an important issue in aeolian research. Surface characteristics affect sand transport processes, but their effects are not yet fully understood. To provide more data on this subject, we observed sand transport in 16 field experiments above surfaces covered by fine, medium and coarse sand. The sand transport rate over relatively coarser‐grained medium and coarse surfaces could be expressed as a Gaussian peak function: q _z = a + b exp (?0.5[(|z – C _h|)/d ]^e), where q _z is the measured sediment transport at height z above the bed and a , b , C _h, d , and e are regression coefficients. The measured sand transport flux peak values (H _h) were linearly related to C _h, and both values were significantly related to the mean surface grain size. However, for the relatively finer‐grained medium and fine sand surfaces, the sediment transport could be expressed as an exponential function. The cumulative sand transport below 0.1 m was directly related to the mean surface grain size, and the relationship could be expressed as the following exponential function: C _z = f + g exp –M _z/i , where C _z is the cumulative sand transport at height z above the bed, M _z is the mean grain size and f , g , and i are regression coefficients. Above 0.1 m, there were no significant relationships between the cumulative sand transport and the mean surface grain size. The mean grain size decreased with increasing height below the peak height and then increased with increasing height. The surface grain size distribution and proportions of the particles in different grain size categories controlled the mean grain size as a function of height. The observed changes in the sand transport rate and grain size with height will provide support for sand disaster mitigation, numerical modelling and studies of dune formation. Copyright © 2016 John Wiley & Sons, Ltd.     

基于分数阶小波域GSM模型的地震信号随机噪声压制方法

地震信号中的随机噪声是一种干扰波,严重降低了地震信号的信噪比,并影响着资料的后续处理和分析.本文根据地震信号中有效信号和随机噪声的差异,结合分数阶B样条小波变换与高斯尺度混合模型提出了一种地震信号随机噪声压制方法.首先利用分数阶B样条小波变换将含噪地震信号映射到最优分数阶小波时频域内,然后对各小波子带系数分别建立高斯尺度混合模型,由贝叶斯方法估计出源地震信号小波系数,最后使用分数阶B样条小波逆变换重构得到降噪后的地震信号.利用本文方法对合成地震记录和实际地震信号进行降噪处理,实验结果表明本文方法能够有效地压制地震信号中的随机噪声,并且较好地保留了有效信号.     

On the prediction of persistent processes using the output of deterministic models

A problem frequently met in engineering hydrology is the forecasting of hydrological variables conditional on their historical observations and the hindcasts and forecasts of a deterministic model. On the contrary, it is a common practice for climatologists to use the output of general circulation models (GCMs) for the prediction of climatic variables despite their inability to quantify the uncertainty of the predictions. Here we apply the well-established Bayesian processor of forecasts (BPF) for forecasting hydroclimatic variables using stochastic models through coupling them with GCMs. We extend the BPF to cases where long-term persistence appears, using the Hurst-Kolmogorov process (HKp, also known as fractional Gaussian noise) and we investigate its properties analytically. We apply the framework to calculate the distributions of the mean annual temperature and precipitation stochastic processes for the time period 2016–2100 in the United States of America conditional on historical observations and the respective output of GCMs.